It’s May 1st.

Besides marking the two week countdown until spring semester ends, it’s also when I get serious in thinking about what the goals for the summer build season are. Historically, the summer season has been spent preparing the robots for Dragon*Con in September. Now, robots are fun and all, but I think they have become routine. If nothing else, taking on a more in-depth project is just a means of self-improvement. Last summer, I also completed and refined LOLrioKart, which gave me a taste for larger-scale, more involved projects. Overcoming its inherent Course VI difficulties taught me a great deal of power electronics and controls knowledge. Before that, RazEr and its predecessors were built to explore electric motor theory and construction. Segfault was built is being built (It’s under about 50 pounds of other stuff, but it’s coming, I PROMISE) to explore digital feedback controllers. So, with every project, I try to do something new or out of the ordinary that I haven’t done before. In the past, this has been limited to trying out new robot designs, but with my increased resource access, I’ve been able to wander outside that domain.

With increased scope and complexity comes the tallest mountain of cruft to sort through – increased cost. As much as LOLrioKart was scrounged, cursory appraisal of all its materials and parts, not counting those that have been scrapped or replaced, approaches the $1000 mark. This isn’t even including the $600+ Briggs and Stratton ETEK motor, which you can’t even get any more, and which I loaned from a friend. Nor does it factor in the price of the batteries if they were purchased new, or really any of the support equipment.

I can’t exactly fund this kind of stuff with a part-time UROP, and scrounging only goes so far before I start rejecting the compromises I have to make in order to get something resembling the original plan through and realized.

Super LOLrioKart

So with this in mind, I applied to MIT’s Eloranta Summer Research Fellowship in the name of great justice being able to realize my latest and greatest bad idea. I have always been a vocal supporter of a student project  fund that undergraduates can apply to in order to get resources for the scientific or engineering initiatives of their own. The Eloranta grant seemed the closest fit to this ideal. And it was $6,000. That’s like, a 6 with 3 zeroes after it, or about how much money the Federal Government spends every 42 microseconds.

But wait, how the hell can I possibly make LOLrioKart any better? Isn’t it already on the cover of Engineers Gone Wild?

No, it isn’t. It’s an electric gokart that has bonus points for absurdity. But LK doesn’t really do anything new that anyone else with a DC motor can’t pull off. It was a great mental exercise for me when I built the differential, or 5 different iterations of the motor controller, but that’s about it. It doesn’t even have supercapacitor boost.

The original plan for Super LOLrioKart dates back as far as the LK build itself. The idea was to have four 16kW custom hub motors, for a total of 64kW of peak motor power and the ability to drag race Tesla Roadsters. The project was then known as LOLrioKart 64. If you think about it, the power to weigh ratio of a < 300 pound kart with 64kW of motor power is not exactly trivial. But 64,000 watts is almost 90 electric horsepower, and the controls for such power levels are not exactly trivial either. Knowing my luck with LK’s DC motor controller, I wasn’t about to test it on high powered 3 phase ones; and so LK64 was shelved before I had even modeled up a hub motor.

It took a little bit more work at the Media Lab before I latched onto my next idea. Let’s say that it’s the common solution point of a CityCar, a swerve drive, and a LOLrioKart. I had an agenda to pursue with this one – I had become tired of the fussy wireless joystick derived controller for the test car. Seriously, it’s a car. You’re supposed to drive it like a car, not like a flight simulator. I set out to design a better interface for the test cars, wireless or otherwise, which were based on real car controls – steering wheel, pedals, shifters, some buttons.

The idea still involved 4 hub motors, but they would be more vestigial and just provide a level of reasonable maneuverability for the vehicle. No, see, the real focus of Super LOLrioKart would have been the four independently steered wheel pods.

Inside each pod is a ~1.5kW hub motor driving the 10″ tires (same size that LK has now) and a stock 700-size motor gearbox that handles the steering. The whole assembly would pivot around the large center gear, which is fixed to the vehicle through the upper half of each wheel arm, only partially modeled here. Each pod was designed to be independent, having internal computation and motor drivers running both drive and steering in closed-loop mode, and would only need a power connection and a point of reference. The control topology itself would be fully a wireless star network based on XBEE radios, which are extremely popular in the DIY electronics world and well-documented. Because of the full 360 degree turning ability of each wheel, there was no way you could convince me to run signal cables to each controller. It was the next stage after drive-by-wire – drive-by-wireless.

And so I wrote a 15 page proposal exposing the control agenda and some of the engineering details of this vehicle, including a rough cost breakdown based on glances at parts prices, and “timeline mitigation factors” i.e. answers to the question “Why the hell would we believe that you can build a CAR in 3 months?”

I mean, I built this in like…. a day and a half. Does that count?

…

They didn’t.

Well, I guess I can find solace in the thought that maybe it was so awesome and over the top that it wrapped back around to the other end, and thus was denied. But actually not – I probably shouldn’t have mentioned the whole ‘shopping cart’ dealie, because that would have passively thrown it from the realm of research and development work – legit stuff to ask a few thousand dollars for – to me punting a summer away by building something epic.

Which, while I consider a fully legitimate reason, obviously a group of people I have never met could not be persuaded to believe. There is something to be said about my proposal-writing and persuasion abilities if that was the case. At any rate, design of SLK stopped on the same day. It was simply going to cost too much for me to think about pursuing without free money, at least at the time of last ponderance.

But would it actually? And what are the tradeoffs I would have to make? I went back and looked at the “first order bill of materials” I had put together in accordance with the recommended format.

Many lines out of that BOM were assumed “commercial purchases”. Stock parts, things I could make or find a friend to make, but would purchase just because I had the funds, in order to save time. The first thing to get cut was the 40Ah LiFePO4 battery. MIT EVT still has their stock of small format cells that I could tap or beg. SLK would a minimum of 48 volts and 20AH, an arrangement easily supplied by a 15S10P A123 array. Many other things were gross overestimates or rough SWAGs. I can’t even think of a way to spend $400 on driver controls. Even a Logitech Driving Force wheel only costs $150 and would be ideal with a little Arduino fiddling.

The $1,400 of motor controllers was estimated from looking at what options Kelly Controller has to offer. As far as I know, nobody even makes a 48 volt DC motor controller small enough to fit inside the wheel pod. So at least some of the controls would have to be custom – and it was a problem that I seeked an answer from someone who knows alot more about motor controllers than I care to think about.  Hopefully, it would be solvable for far less than $1,400 (but likely not that much less) and some programming. Oh god the programming.

I figured there was no way I was going to escape from the raw material cost since I’d need serious metal to build the wheel modules. Also, the sheer volume of NdFeB in the motors would not cause the cost to wander too far from $400 even if I used stock flat magnets. Practically all the big power electrical components I either already have or have access to, so the “Power bus” is negligible.

With adequate scrounging and borrowing, I could probably bring the out of pocket cost under $3,000, with strong tradeoffs in commercially made parts towards DIY and fabrication. This would probably then break the project out of its summer-only timeline, too. On top of that, summer housing and living expenses take #1 priority, since exercising shopping cart absurdism is a little too close to being a hobo than I feel comfortable with.  It’s still financially unreasonable.

It’s a project that I think has the right about of Incremental Epicness given last summer’s work, which is why I haven’t dropped the idea completely. And the agenda – of course, the agenda.You don’t need joysticks to control an omnidirectional car, and I’m want to prove it.

But at the same time, I thought that perhaps I should just start on a different path completely, something that doesn’t require expensive power electronics and wireless embedded networks and lithium batteries.

Read more “The Summer Build Season 2010: Super LOLrioKart and Chuckranoplan”